iRule sorry/maintenance page on pool with one node

When you restart the pool member (server), it is almost certain that the health monitor that is checking that member will fail and mark the pool member as down. The monitor down takes precedence over the manual disable and for good reason. The BIG-IP system does not want to send any traffic to a pool member that is truly down (e.g. being restarted). When you manually disable a pool member, new connections are not allowed, but persistent connections are. If the server is restarting, you do not want to direct a persistent connection to the server; you want the connection load balanced to another server. Manual disable is designed to gracefully remove traffic from a particular server so that it can eventually be taken offline for maintenance. But offline, whether by a monitor or a manual Force Offline, results in the same pool member LB::status from the iRule's perspective: down.